zonal: make apply() output .name deterministic across backends

[brendancol]

Closes #2611

What changed

• zonal.apply() now sets a deterministic .name on its output and accepts a name parameter (default None), matching regions, hypsometric_integral, trim, and crop.
• The name is assigned after the DataArray is built. xr.DataArray(dask_array, name=None) otherwise inherits the dask graph's internal task name, which differs from the numpy/cupy backends and changes between runs.

Backend coverage

numpy, cupy, dask+numpy, dask+cupy. The new test runs on all four; before this change the two dask backends returned a task-hash name while numpy/cupy returned None.

Test plan

• test_apply_name_consistent_across_backends (new): default name is None and an explicit name= is honored on every backend
• Full zonal suite: 213 passed
• flake8 clean on xrspatial/zonal.py and the added test lines

[brendancol]

PR Review: zonal: make apply() output .name deterministic across backends

Blockers (must fix before merge)

None.

Suggestions (should fix, not blocking)

None.

Nits (optional improvements)

• The new name parameter goes at the end of the signature, after rasterize_kw. That keeps existing positional calls working and matches how regions/trim/crop expose name, so it is fine as-is. Noting the placement was a deliberate compatibility choice.

What looks good

• Root cause is correct: xr.DataArray(dask_array, name=None) inherits the dask graph's task name, so assigning .name after construction is the right fix. Setting it at construction time does not override the dask name.
• The test runs on all four backends and checks both the default (None) and an explicit name, which is exactly the inconsistency being fixed.
• Attrs, dims, and coords handling is untouched; only .name changed.
• The code comment explains why the assignment happens after construction, which is the non-obvious part.

Checklist

• Algorithm matches reference/paper: n/a (metadata fix)
• All implemented backends produce consistent results: yes, verified by the new parametrized test
• NaN handling is correct: unchanged by this PR
• Edge cases covered by tests: name default and explicit name both tested
• Dask chunk boundaries handled correctly: unchanged
• No premature materialization or unnecessary copies: confirmed, no compute added
• Benchmark exists or is not needed: not needed
• README feature matrix updated: not applicable, no new function
• Docstrings present and accurate: the new name parameter is documented